Caribbean Current Variability and the Influence of the Amazon And

ARTICLE IN PRESS

Deep-Sea Research I 54 (2007) 1451–1473 www.elsevier.com/locate/dsri

Caribbean current variability and the inﬂuence of the Amazon and Orinoco freshwater plumes

L.M. Che´rubina,Ã, P.L. Richardsonb

aRosenstiel School of Marine and Atmospheric Science, 4600 Rickenbacker Causeway, FL 33149 Miami, USA bDepartment of Physical Oceanography, MS 29, Woods Hole Oceanographic Institution, 360 Woods Hole Road, Woods Hole, MA 0254, USA

Received 5 February 2006; received in revised form 16 April 2007; accepted 24 April 2007 Available online 18 May 2007

Abstract

The variability of the Caribbean Current is studied in terms of the influence on its dynamics of the freshwater inflow from the Orinoco and Amazon rivers. Sea-surface salinity maps of the eastern Caribbean and SeaWiFS color images show that a freshwater plume from the Orinoco and Amazon Rivers extends seasonally northwestward across the Caribbean basin, from August to November, 3–4 months after the peak of the seasonal rains in northeastern South America. The plume is sustained by two main inflows from the North Brazil Current and its current rings. The southern inflow enters the Caribbean Sea south of Grenada Island and becomes the main branch of the Caribbean Current in the southern Caribbean. The northern inflow (141N) passes northward around the Grenadine Islands and St. Vincent. As North Brazil Current rings stall and decay east of the Lesser Antilles, between 141N and 181N, they release freshwater into the northern part of the eastern Caribbean Sea merging with inflow from the North Equatorial Current. Velocity vectors derived from surface drifters in the eastern Caribbean indicate three westward flowing jets: (1) the southern and fastest at 111N; (2) the center and second fastest at 141N; (3) the northern and slowest at 171N. The center jet (141N) flows faster between the months of August and December and is located near the southern part of the freshwater plume. Using the MICOM North Atlantic simulation, it is shown that the Caribbean Current is seasonally intensified near 141N, partly by the inflow of river plumes. Three to four times more anticyclonic eddies are formed during August–December, which agrees with a pronounced rise in the number of anticyclonic looper days in the drifter data then. A climatology-forced regional simulation embedding only the northern (141N) Caribbean Current (without the influence of the vorticity of the NBC rings), using the ROMS model, shows that the low salinity plume coincides with a negative potential vorticity anomaly that intensifies the center jet located at the salinity front. The jet forms cyclones south of the plume, which are moved northwestward as the anticyclonic circulation intensifies in the eastern Caribbean Sea, north of 141N. Friction on the shelves of the Greater Antilles also generates cyclones, which propagate westward and eastward from 671W. r 2007 Elsevier Ltd. All rights reserved.

Keywords: Carribean; Amazon; Orinoco; Salinity; Drifter; MICOM; ROMS; CDOM; Anticyclone; Cyclone; Instability; Numerical modeling

ÃCorresponding author. E-mail address: [email protected] (L.M. Che´rubin).

1. Introduction the equator in the North Brazil Current (NBC) and flows northwestward along the continental margin River discharge plays an important role in the of South America in the form of a coastal current hydrological cycle and thermodynamic stability of (Candela et al., 1992; Dessier and Donguy, 1994), in the ocean. Knowledge of the variations in the extent NBC rings (Johns et al., 1990; Fratantoni et al., and dispersal patterns of major river plumes and 1995; Gon˜i and Johns, 2001), and as Ekman their mixing rates with oceanic water is critical in all transport in the ocean interior (Mayer and Weis- aspects of continental shelf and regional oceano- berg, 1993). graphy. In particular, freshwater seasonally im- During the summer and fall each year a major pinges on coral reef ecosystems, which affects the part of the NBC retroflects near 61N and feeds into recruitment of larval reef fish (Kelly et al., 2000; the eastward-flowing North Equatorial Counter- Cowen and Castro, 1994) and fish mortality (Hu current which is intensified then. In spring, the et al., 2004). The Caribbean Sea is influenced by the Countercurrent weakens and reverses at the surface dispersal of the freshwater from the Amazon and as a result of westward Ekman flow. Occasionally, Orinoco Rivers, which is discharged into the pieces of the NBC retroflection pinch of as large tropical Atlantic and advected into the Caribbean NBC rings (around 400 km in overall diameter), Sea (Fig. 1). which translate northwestward toward the Carib- The Caribbean Current is a major current, which bean Sea (Johns et al., 2003; Garzoli et al., 2003; transports South Atlantic water through the Car- Gon˜i and Johns, 2003; Fratantoni and Glickson, ibbean and into the Florida Current and the Gulf 2002; Ffield, 2005; Frantantoni and Richardson, Stream. It is an important conduit of the upper part 2006). Recent results from the NBC Rings Experi- of the northward-flowing meridional overturning ment suggest that 8–9 rings form per year transport- circulation (Schmitz and Richardson, 1991; Schmitz ing roughly 9 Sv (Johns et al., 2003) with no marked and McCartney, 1993). South Atlantic water crosses seasonal variability of the formation rate but with

25oN

−1000

20oN Puerto−Rico − Hispanola 1000 − 1000 Leeward −1000 Islands Mona Island Guadeloupe

15oN Caribbean Sea

St Vincent Windward Islands Atlantic Ocean −1000 Grenada Passage

o 10 N NBC rings Orinoco −1000 Retroflection North Equatorial Countercurrent 5oN

− 1000 North Brazil Amazon Current 0o

5oS

80oW 75oW 70oW 65oW 60oW 55oW 50oW 45oW 40oW 35oW

Fig. 1. Map of the Caribbean Sea and of the nearby tropical Atlantic showing the Amazon and Orinoco rivers along with the various groups of islands cited in the text. ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1453 seasonal variations of ring structure (see Johns and low salinity 34.5 to the north. Low salinity et al., 2003) and, possibly, significant year-to-year values near 35.5 in the upper 50 m extended north to variability (Gon˜i and Johns, 2003). Some rings around 181N, indicating the freshwater plume disappear after they collide with the continental passed northward through the eastern Caribbean margin and islands south of 141N. Most rings at that time. Using the same Regional Ocean translate northward as they approach the Lesser Modeling System (ROMS) simulation as in this Antilles. They often interact and merge with each study, Baums et al. (2006) showed that intensifica- other around 141–181N, where they tend to stall and tion of the jet south of the domain (14–151N) was decay (modeling study by Garraffo et al., 2003; followed by an increase of the presence of mesoscale Frantantoni and Richardson, 2006). No ring has anticyclones south of Puerto Rico and Hispanõla. been observed to coherently enter the Caribbean Baums et al. (2006) focused on the impact on coral through the island passages, although a modeling reef larvae of the seasonal changes of mesoscale and study by Simmons and Nof (2002) suggests that this small-scale eddy activity due to the presence of the is possible for large rings encountering the smaller freshwater plume. To do so, they defined two islands to the north of the Grenadines. We infer seasons, January–July (no freshwater advected into from the available results that a main source of the Caribbean Sea) and August–December (plume South Atlantic water flowing into the Caribbean Sea of freshwater advected into the Caribbean Sea). north of 141N is carried to the islands by rings. They showed that, during the spawning season, in Johns et al. (2002) showed that the overall August, larvae would become trapped in small distribution of the Atlantic inflow into the Car- topographically steered eddies near the Mona island ibbean is nearly evenly divided among three main between Puerto Rico and Hispanõla (Fig. 1). No groups of passages: the Windward Islands passages, exchange of larvae would occur then between the the Leeward Islands passages and the Greater eastern and western Caribbean. If larvae were Antilles passages. Most of the inflow through the released later in summer or fall, then dispersion Windward Islands passages occurs through the would occur due to the presence of the anticyclones Grenada Passage located south of Grenada and hugging the southern coast of Puerto Rico and near 11.51N(Johns et al., 2002), forming the Hispanõla. Such dispersal variability was substan- main Caribbean Current. Johns et al. (2002) report tiated by genetic analysis of the coral larvae. mean transports of 5.770.8 Sv through Grenada This study focuses on how seasonal variations of Passage (11.71N), 2.970.8 Sv through St. Vincent the freshwater plume change the velocity structure Passage (13.51N), and 1.570.8 Sv through St Lucia and the meridional extent of the Caribbean Current Passage (14.31N), where 1 Sv ¼ 106 m3/s. The inflow in the eastern Caribbean Sea. We address the link is fed by the main NBC and from NBC rings that between the seasonal supply of freshwater from the collide with the continental margin near Tobago Orinoco and Amazon rivers and the two inflows, (near 11.2N). Two bands of westward flow were which transport this freshwater into the Caribbean observed by drifters in the southeastern Caribbean Sea. We focus on the sources of the Caribbean (611–641W), one located near 11.51N and the other Current that passes through the Windward Islands farther north near 141N(Centurioni and Niiler, passages, south of Guadeloupe (16.31N–61.51W). 2003; Richardson, 2005). The northern band ap- We demonstrate that the intensification of the pears to originate in the flow through St. Vincent Caribbean Current near 141N is correlated with and St. Lucia Passages. The two bands merge the arrival of the freshwater plume as observed with downstream in the Caribbean Current near 661W sea surface salinity (SSS) maps, surface drifters and, (Richardson, 2005). Sea-viewing Wide Field-of-view Sensor (SeaWiFs) Intensification of the Caribbean Current asso- colored dissolved organic matter (CDOM) maps. In ciated with the Orinoco freshwater plume was order to understand the intensification of the shown by Hernandez-Guerra and Joyce (2000).In Caribbean Current by the freshwater inflow and September 1997 they observed a fast (130 cm/s) the origin of the Caribbean anticyclones, we used westward jet centered at 131N and 661W near the two numerical simulations. The first one is a southern edge of a mass of relatively freshwater with realistic high-resolution (1/121, i.e. 8 km grid salinities around 34.5 located between 13–141N. The spacing) numerical simulation from 1979 to 1986 jet coincides with the salinity gradient region of the North Atlantic Ocean with the Miami located between high salinity 37.0 to the south Isopycnic Coordinate Ocean Model (MICOM, ARTICLE IN PRESS 1454 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

Bleck et al, 1992; Bleck and Chassignet, 1994), NBC retroflection and North Equatorial Counter- which has outputs available online at http:// current, leaving a remnant plume northwest of the hycom.rsmas.miami.edu. This simulation is used retroflection. By December the eastward plume in to verify and understand the links found in the the countercurrent has weakened and the earlier observations between the freshwater plume and the remnant plume northwest of the retroflection changes of the Caribbean Current dynamics. In this appears to have drifted into the Caribbean. In simulation both the influence of the NBC rings and August at the time of the Orinoco’s maximum the freshwater plume are present. In order to test discharge (0.07 Sv, Hellweger and Gordon, 2002), the unknown contribution of NBC rings and the its freshwater plume extends northward and merges influence alone of the freshwater plume to the with the remnant Amazon plume. Between August Caribbean Current and anticyclones, a second and November the Orinoco plume appears to pass regional simulation, without NBC rings is carried northwestward through the eastern Caribbean, on with the ROMS (Shchepetkin and McWilliams, contributing significant freshwater, which merges 2005) model. This simulation encompasses only the with Amazon water there. In the southeastern northeastern Caribbean Sea so that NBC rings are Caribbean salinities in the Orinoco plume drop not resolved, and it is forced by Levitus climatology below 34.0, reaching 33.0 in September. The (Da Silva et al., 1994). The freshwater plume is seasonal variation of salinity in the eastern Car- present and is shown to be the main source of ibbean near 151N651W is shown in Fig. 3. negative potential vorticity anomaly (PVA) in the Although the Amazon’s mean discharge is northeastern Caribbean, which strengthens the around 6 times that of the Orinoco (Perry et al., Caribbean Current in that region. 1996), the Amazon water is dispersed over a much larger area and travels nearly 2000 km to reach the 2. Observations Caribbean; the Orinoco River is only 300 km from the Caribbean and arrives there quickly (10 days 2.1. Salinity maps at an estimated 30 km/day), less diluted than the Amazon plume. Therefore, the Orinoco freshwater The salinity maps were downloaded from the plume appears to be dominant in the eastern 2003 monthly climatology of the tropical Atlantic Caribbean during August–November and to cause sea-surface salinity (401S–601N).1 The data used in the lowest salinities there. The evidence for this is the salinity maps were collected from ships of that the Orinoco plume in August–September is opportunity for the French institute for research separated from the main Amazon plume by a gap of and development (IRD) and from the world ocean higher salinities. However, the Orinoco plume is database 2001 (WOD01) of the national oceano- added to the remnant Amazon plume in the graphic data center (NODC). The monthly maps Caribbean and the seasonal pattern of Orinoco were generated from 11 11 spatial averages and discharge is similar to the 2-month lagged Amazon objective analysis of all the data. Further details on discharge, which makes it difficult to distinguish the the data processing and analysis are given by relative contributions to the eastern Caribbean Dessier and Donguy (1994). (Hellweger and Gordon, 2002). The salinity maps (Fig. 2) show seasonal variations of the two freshwater plumes that originate in 2.2. Ocean color maps the Amazon and Orinoco River discharges. The Amazon plume extends northwestward toward the Hu et al. (2004) studied the correlation between Caribbbean during January–June while the salinities CDOM and SSS comparing SeaWiFS CDOM in the Caribbean remain fairly high then (36.0). At maps with salinity profiler floats (S-PALACE). the Amazon’s maximum discharge in June (0.23 Sv, They showed monthly maps of ocean color, which Muller-Karger et al., 1989; Hellweger and Gordon, can be interpreted to describe details of the 2002), its freshwater plume (o34.0) covers a wide freshwater plumes during 5 years (Fig. 4), and area of the tropical Atlantic from 0–151N and they indicated that most of the freshwater plume 50–601W. Starting in July the main Amazon in the western tropical Atlantic originates from discharge appears to be advected eastward by the the discharge of the Amazon River. Moreover, they showed that the freshwater plume in the 11 http://www.brest.ird.fr/sss/clim_atl.html Caribbean Sea originates from both the Amazon ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1455

Fig. 2. Monthly climatology surface salinity maps of the western tropical Atlantic and eastern Caribbean from the French Institute of Research and Development (IRD). The maximum discharge of the Amazon River (0.23 Sv) occurs, on average, in June; maximum Orinoco River discharge (0.07 Sv) is in August (Muller-Karger et al., 1989; Hellweger and Gordon, 2002). ARTICLE IN PRESS 1456 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

Fig. 2. (Continued) ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1457

36.2 122 westward in the Caribbean merging with the 115 36 112.2 Amazon plume.

35.8 102.4 Maps of CDOM distribution given by Hu

35.6 92.6 et al. (2004) (see also Muller-Karger et al., 1989) show very clearly the advection of the fresh- 35.4 82.8 water plume by NBC rings (Fig. 4). We compare 35.2 73

here the maps of CDOM in May 1998 with Salinity 35 63.2 the identiﬁcation of NBC rings by Fratantoni 34.8 53.4 and Glickson (2002). Their NBC ring D passed

34.6 43.6 551W on May 4, 1998, which corresponds to O−Salinity 34 the patch of high CDOM concentration just 34.4 x−Looper days 33.8 east of the Lesser Antilles. Another ring passed 34.2 24 0 2 4 6 8 10 12 14 551W in July 1998 according to Fratantoni and Month Glickson (2002) and is revealed by the CDOM Fig. 3. Seasonal variations of salinity in the eastern Caribbean. patch southeast of the Lesser Antilles on the Salinity values were obtained from the salinity maps (Fig. 2)in July map. With no other rings formed until after the eastern mid-Caribbean at 151N–651W. Maximum salinity these two, the CDOM tongue dissipated east of the 36.05 occurs in March and minimum 34.45 in September– Lesser Antilles. This scheme is repeated for years October, a salinity range of 1.6. Salinity values in the southeastern corner of the Caribbean near Grenada Island, where the 1999 and 2000 (Hu et al., 2004, their Fig. 2). An Orinoco plume enters, have a larger seasonal variation than this inspection of the monthly maps shows blobs and (2.8). Also shown are seasonal variations of the number of filaments of higher and lower CDOM in the monthly anticyclonic drifter looper days (65–751W), which are Caribbean. These are interpreted to have been interpreted to indicate the population of anticyclones as discussed caused by current filaments and eddies that ad- later (from Richardson, 2005). The looper days were smoothed with a running 3-month average. vected the plume in complicated patterns. The smooth freshwater plume in Fig. 2 is an average of the filaments and blobs. This seasonality of freshwater plumes allowed us and the Orinoco rivers. In July–August, a to determine heuristically two periods of time mixture of the Amazon and Orinoco freshwater according to the salinity content of the surface plumes extends northwestward across the eastern water of the eastern Caribbean Sea as observed in Caribbean. the salinity and ocean color maps (Figs. 2 and 4). The CDOM distribution maps suggest that Lowest salinity in the eastern Caribbean is observed during summer and fall much of the Amazon River in August–October (Fig. 3). In winter there is little plume is entrained into the NBC and flows around freshwater input from rivers so salinity remains the retroflection and into the North Equatorial high. In order to identify changes in the Caribbean Countercurrent (Muller-Karger et al., 1988). During Current dynamics we averaged quantities over two this time, when a large part of the Amazon plume periods of time: the first one, of high salinity, is from extends eastward as demonstrated by SSS maps January to July, and the second, of low salinity, (Fig. 2), NBC rings periodically pinch off from the from August to December. This choice will be NBC and carry entrained Amazon water north- confirmed a posteriori by the results of this study. westward toward the Caribbean (Fratantoni and Such seasonal grouping was done by Ffield (2005), Glickson, 2002). Therefore, rings contribute to a who shows an increase of the historical surface mean northwestward motion of the freshwater temperature of 2.2 1C in the Caribbean Sea along tongue east of the Lesser Antilles and contribute with the low salinity plume. freshwater to the Caribbean between 141 and 181N as they decay there (Frantantoni and Richardson, 2.3. Surface drifter data 2006). The ring contribution would be in addition to any freshwater from the Amazon dispersed into the Between 1996 and 2003, 212 satellite-tracked North Equatorial Current and transported west- drifting buoys measured trajectories and over ward to the Caribbean. The CDOM and SSS maps 73,000 6-h velocities in the Caribbean Sea (Richard- also reveal that the Orinoco freshwater plume son, 2005). The drifting buoy data were acquired passes through the Grenada passage and north- from the Global Drifting Buoy Data Assembly ARTICLE IN PRESS 1458 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

Fig. 4. Monthly composite maps of CDOM distribution estimated from Sea WiFS version 4 data for 1998 reproduced from Hu et al (2004), who showed five years of data (1998–2002). Overlaid on the images as red crosses are concurrent S-PALACE float locations. Red and yellow show the high CDOM concentration advected along the Brazilian and Venezuelan shelves. The retroflection of the NBC and the formation and northwestward translation of NBC rings can be observed from May to December. In the eastern Caribbean the CDOM plume spreads northwestward between July and November corresponding to the low salinity in Fig. 2.

Center at the NOAA Atlantic Oceanographic and Following Richardson (2005) surface drifter Meteorological Laboratory in Miami, Florida. The velocity and eddy kinetic energy (EKE) were majority of the drifters were similar to the averaged in boxes by grouping all available 6-h WOCE–TOGA Lagrangian drifter described by drifter velocities into 1/21 1/21 bins for Januar- Sybrandy and Niiler (1991). Drogues were attached y–July and August–December. In each box, mean below the surface float and centered at a depth of velocity was calculated as the sum of all u (v) 15 m. The typical accuracy of a velocity measure- velocity components in the x (y) direction divided ment in 10 m/s wind is 0.01 m/s (Niiler et al., 1995). by the number of observations. EKE was calculated ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1459 by averaging the u and v velocity variances meridional velocity sections along 63.51W(Johns (variances about the mean velocities) in each box. et al., 1999). The standard error of mean velocity O2st/N, where In the central Caribbean (65–701W, Fig. 6b) there s is the variance of velocity about the mean velocity, are also 3 jets during August–December (red line) N is the number of 6-h velocity observations divided with the fastest (65714 cm/s) located in the south by 4, and t is the integral time scale of the near 11.31N. The central jet (13.81N) is significantly Lagrangian autocorrelation function, was estimated faster (3675 cm/s) than the January–July (blue) to be 2 days. The number of degrees of freedom, velocities (2474 cm/s). During January–July there N/t, was calculated by summing the number of are only two jets as if the central (13.81N) and 2-day intervals spent by a drifter within a box. southern (11.81N) jets in the east (61–641W) had Further information about the data and calculation merged into one near 12.51N (4679 cm/s) in the of velocities is given in Richardson (2005). central region. The main point is that the swift speeds in the central jet (13.81N) during August– 2.3.1. Surface drifter velocity maps December coincide with the inflow of freshwater The main Caribbean Current along the southern into the Caribbean, some of which comes from boundary of the Caribbean appears in both seasons NBC rings that impinge on the islands near as a band of red arrows indicating speeds greater there and stall between 141N and 181N (Section than 25 cm/s (Fig. 5). Also in both seasons two 2.2). This appears to cause a farther westward bands of swift flow enter the southeastern Carib- extension of the 141N jet during August–December bean, a southern one south of 121N and a northern as the salinity front propagates westward. Intensi- one near 141N. Between these two, inflow appears fication of the 141N jet could be caused by the lower to be partially blocked by the line of islands salinity and higher temperature carried by the extending northward from Grenada-Grenadines- freshwater plume. Both increase the speed of the St. Vincent, between 12.01 and 13.31N. During density-driven flow by establishing a stronger August–December the northern band near 141N density gradient at the southern part the freshwater appears to be swifter than during January–July and plume where it lies adjacent to the Caribbean Sea flows westward across the eastern Caribbean, water. merging with the main boundary flow near 661W. The northern inflow near 141N is in the latitude 2.3.3. Caribbean anticyclones range where NBC rings stall and decay, east of the Looping drifter trajectories in the Caribbean were islands, and thus rings contribute to this inflow. In used by Richardson (2005) to identify discrete both seasons part of the inflow near 141N appears eddies, to plot their trajectories, and to determine to be deflected northward (15–161N) east of the eddy characteristics. A looper was defined as a Aves Ridge, which is located near 63.51W. trajectory that contained two or more consecutive loops in the same direction. Box averages of drifter 2.3.2. Caribbean jets data in the central eastern Caribbean (65–751W) The mean meridional structure of the Caribbean revealed that of the 14,196 total number of Current is revealed by averaging drifter velocities in observations, 25% were in loopers and 71% of half-degree latitude bins located in the eastern these were anticyclonic, suggesting that anticyclones (61–641W) and central (65–701W) Caribbean are the dominant eddies and make up 17.5% of the (Fig. 6). In the eastern Caribbean (Fig. 6a), three drifter data. If the percentage of anticyclonic looper jets are observed during both seasons. The swiftest data indicates the percentage of the area consisting is located farthest south near 11.31N. It appears to of anticyclones and their typical overall diameter is be faster (81715 cm/s) during January–July (blue 250 km, then there would be an average population line) than during August–December (5478 cm/s) of 2.2 anticyclones in the box. This agrees with two (red line), where the standard errors are given. It pairs of anticyclones tracked simultaneously with also shifts slightly north during August–December. drifters. At their typical westward translation A central jet (21–35 cm/s) is located near 141N, and velocity of 13.4 cm/s, anticyclones translate through a northern jet (10–15 cm/s) is located near 171N the box in around 3.0 months, which suggests a with somewhat faster speeds during August– formation rate of around 8 anticyclones per year. December (red line). Three similar jets have been Choosing a smaller diameter of 200 km results in observed in this region in an average of five around 12 anticyclones per year. The typical swirl ARTICLE IN PRESS 1460 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

Mean Velocity Vectors Jan-Jul

18°N

15°N

12°N

50 cm/s

75°W 72°W 69°W 66°W 63°W

Mean Velocity Vectors Aug-Dec

18°N

15°N

12°N

50 cm/s

75°W 72°W 69°W 66°W 63°W

Fig. 5. Mean surface velocity vectors calculated by grouping 6-h drifter velocity values into 1/21 1/21 bins (as in Richardson, 2005). Vectors are shown for two time periods, January–July (a) and August–December (b). Red arrows indicate speeds greater than 25 cm/s. Vectors are shown in bins that contained two or more degrees of freedom, based on a 2-day integral time scale of the Lagrangian autocorrelation function, which was estimated from the drifter velocity series. Typical standard errors of the mean eastward velocity components are around 16 cm/s. Depth contours are 200 m (dashed) and 2000 m (dotted). ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1461

Surface Drifter Eastward Velocity (61−64W) 18

17 Aug−Dec

Jan−July 16

14 Latitude 13

10 −100 −80 −60 −40 −20 0 20 U Velocity (cm /sec)

Surface Drifter Eastward Velocity (65−70W) 18

16 Aug−Dec Jan−July 15

14 Latitude 13

10 −100 −80 −60 −40 −20 0 20 U Velocity (cm/sec)

Fig. 6. Profiles of mean eastward velocity and standard error calculated by grouping surface drifter velocities in 1/21 latitude bins by season. (a) 61–641W zonal average in the eastern Caribbean. (b) 65–701W zonal average velocity in the mid-Caribbean. Blue profiles are for January–July and red profiles for August–December. velocities in the anticyclones are around 60 cm/s, current filaments and eddies that advect and stir the and their diameters are around 200 km, as indicated freshwater plumes in the Caribbean. by looping drifters. The loopers and other compli- The number of anticyclonic looper days per cated drifter trajectories give some insight into the month in the 65–701W box varies seasonally, with ARTICLE IN PRESS 1462 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 high values reaching 115 days during October, and (ECMWF) daily atmospheric data from 1979 to low values dropping to 34 days during March 1986 (Chassignet and Garraffoo, 2001). (Fig. 3). This suggests a seasonal variation in the The simulation reproduces the most important number of Caribbean anticyclones that could be characteristics of flow in the Caribbean (Romanou related to variations in the velocity of the Caribbean et al., 2004). The total model transport into the Current or variations in the structure of NBC rings, Caribbean Sea through the Windward and Leeward which collide with the Antilles (Richardson, 2005). Islands Passages is 25 Sv, well within the observa- However, Fig. 3 shows that the seasonal variation tional estimates of 18–33 Sv. The mean Florida of anticyclonic looper days is negatively correlated Straits transport in the model is 31.4 Sv; values peak (0.9) with the salinity in the eastern Caribbean, or in the summer months at the observed transport and positively correlated (0.9) with the amount of with the correct seasonality, maximum in July and freshwater in the Caribbean. The implication is that minimum in October (Richardson et al., 1969; more anticyclones are generated during the time of Schott et al., 1988; Molinari, 2004; Hamilton the freshest water, nearly twice as many during et al., 2005). Model NBC rings provide 40% of August–December (95 looper days per month) as the water for the Caribbean transport (Johns et al., during January–July (51 looper days per month). 1998, 2002), and the simulated rings have three The modeling studies reported below suggest a different structures as observed: some are surface dynamical connection between the freshwater plume intensified, some subsurface intensified with a sur- and the anticyclones, which helps explain these face signature and some subsurface intensified with results. no surface signature (Garraffo et al., 2003; Fran- tantoni and Richardson, 2006). The generation rate 3. The Caribbean current in numerical models for simulated rings is 7–9 per year, of which 6 are surface-intensified, in good agreement with altime- In this section two numerical simulations are used try (Gon˜i and Johns, 2001). The model also displays to verify the links between the presence of the strong mesoscale variability in inter-island passage freshwater plumes, the intensification of the center transports with no clearly defined seasonality, in jet of the Caribbean Current, and the presence of good agreement with observations. Furthermore, anticyclones in the northeastern Caribbean. First a the model EKE in the Caribbean during the 6-year North Atlantic global simulation is compared to spinup period averaged 800 cm2/s2, ranging between previous observations. Then the mechanisms in- 450 and 1250 cm2/s2 (Garraffo et al., 2001); this volved in the links are suggested. Finally, the second agrees well with recent studies by Fratantoni (2001) simulation from a regional model sheds light on the and Richardson (2005), whose calculations using role of the freshwater plume alone (without the surface drifters showed the Caribbean EKE to range effect of the NBC rings vorticity) on the dynamics from 500 to 1500 cm2/s2. Therefore, we have some of the northeastern Caribbean basin. confidence that this basin-scale numerical simulation with high resolution reproduces the large-scale 3.1. MICOM simulation variability as well as the mesoscale variability of the Caribbean Sea. This study uses the high-resolution North Atlan- Surface salinity in the MICOM simulation was tic MICOM simulation, with resolution of 1/121 relaxed to monthly surface salinity from Levitus (mesh size on the order of 8 km on average). The climatology (Da Silva et al., 1994). Freshwater model domain is the North and Equatorial Atlantic fluxes from the Amazon (0.19 Sv) and Orinoco (0.09 (281S–701N), including the Mediterranean Sea, with Sv) rivers were kept constant. Therefore, the 20 layers in the vertical. The ocean boundaries at MICOM simulation has the seasonal variation of 281Sand701N were treated as closed but were the surface salinity driven by the Levitus climatol- bordered by 31 buffer zones in which temperature ogy. However, the salinity variation range remains and salinity are linearly relaxed toward their smaller than in the climatology itself. seasonally varying climatological values (Levitus, From the ECMWF MICOM simulation we 1982). After a 6-year spinup with monthly climato- computed the 1/21 latitude bin velocity averages in logical forcing, the model was integrated using two meridional bands (66–701 and 61–661W) for the surface boundary conditions based on European two seasons, January–July and August–December center for medium-range weather forecasts (Fig. 7). The different meridional bands were chosen ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1463

Fig. 7. Profiles of mean eastward surface velocities (solid line) by averaging velocities in 1/21 latitude bins by season from the MICOM simulation. (a) 61–661W zonal average in the eastern Caribbean. (b) 66–701W zonal average velocity in the mid-Caribbean. Blue profiles are for January–July, and red profiles for August–December. Dashed lines show the standard deviation for each profile. to recover the observed Caribbean current struc- 6 cm/s then). From the drifter data, only the center ture. In the MICOM simulation 61–661W zonal jet is faster during the second season (Fig. 6a). In average the triple jet structure of the Caribbean the western band (66–701W, Fig. 7b), the northern Current is present (Fig. 7a) in both seasons as in the jet centered at 17.51N is faster (15 cm/s) during the observations. The southern jet at 121N is faster than first season as in the drifter data. The center the central one at 151N in both seasons. However, and southern jets merge into one jet centered at the swiftest velocities are obtained during the 141N in both seasons. This is different from the second season for both jets (45 cm/s for the southern observations, which show such a merging only jet and 26 cm/s for the center jet, which is faster by during January–July. An increase of the speed of ARTICLE IN PRESS 1464 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

O(standard deviation) of the center jet in August–- 3.1.2. The anticyclone field December is obtained by calculating the velocity We present in Fig. 9 the rate per month of modulus instead of the zonal velocity as the center anticyclones in the MICOM simulation for both jet flows northwestward (Fig. 8b). seasons. In the mid-Caribbean, around 151Nand 66–701W, the number of anticyclones increases from 3.1.1. Velocity vector map 0.05 to 0.3 per month during August–December From the ECMWF MICOM simulation we also (Fig. 9a,b). In the eastern Caribbean, the eddy field is computed the average vector maps of the eastern almost constant over the two seasons west (leeward Caribbean for the two seasons, January–July and side) of the Lesser Antilles. The westward paths of August–December (Fig. 8). During the first season the anticyclones suggest the role of island wakes in the swift velocities of the southern jet extend up to the eddy formation. The number of anticyclones 141N(Fig. 8a), whereas they are confined south of increases significantly south of Puerto Rico and 131N in the drifter data (Fig. 5a). A northwestward Hispanõla during August–December, which contri- discontinuous flow starting at 141Nand621Wis butes to a slowing of the northern jet (171N). More present in the drifter velocity vectors during anticyclones are generated during the second season January–July (red arrows). It forms an intense flow in the model on the northern side of both the south of Puerto Rico, which is also present in the southern and center jets (Fig. 9b). This agrees well MICOM simulation (Fig. 8a). However, in the with the increase of the number of looper days in the latter the flow at 141N is directed more to the west drifter data (Fig. 3) during the second season, which and at 151Nand651W deviates to the north. It double as in the MICOM simulation. forms the northern jet depicted in the zonal average In order to understand the influence of the on Figs. 5b and 7b. In the drifter data another freshwater on the dynamics of the Caribbean secondary branch flowing northwestward separates Current in the eastern Caribbean we averaged from the weak center jet at 671W and reinforces the spatially the salinity and integrated spatially and northern jet to the west of the eastern basin. The vertically the kinetic energy of the mean flow northern jet flows from 641 to 751W in the drifter (MKE), the eddy kinetic energy (EKE), and the data and from 651W to the end of the domain in the available potential energy (APE) in the box model data. This difference justifies the distinct 61–701W and 13–181N. MKE is obtained from the averaging region used for the model and the spatially weighted averaged flow field, EKE is made observations, and as a result the same jet structure from the departure of the flow from the temporal is seen in both the model and the observations. In average of the flow field, and APE is made from the the MICOM simulation the northern and southern departure of local layer thickness from the temporal jets are separated by a dipolar recirculation cell not local average of layer thickness. The respective time visible in the drifter data. The center jet in the series are plotted in Fig. 10 for the time of maximum MICOM simulation is better defined than in the EKE, July–December. The salinity varies in that drifter vectors. region from a minimum of 34.72 in September to The vector map averaged over the second season 35.65 in February–March (not shown) (Fig. 10). As (August–December) differs from the previous one the salinity decreases in July–September, the EKE through the northwestward orientation of the and APE increase. This increase could be due to the southern jet and the absence of an extended intensification of the Caribbean Current at 141N northern westward flowing swift jet south of Puerto (Hernandez-Guerra and Joyce, 2000) as the hor- Rico (Fig. 8b). In the drifter vectors, the northern izontal shear increases as a result of the strengthen- westward jet is still present but is less intense. The ing of the halocline front and the deepening of the center jet in the drifter data is also more intense and upper layers as the freshwater plume replaces flows westward at 141N all across the eastern Caribbean waters. In September–October APE Caribbean basin. Overall, the northwestward decreases while EKE remains high. This particular drift is emphasized in the MICOM simulation at phase could correspond to the growth of baroclinic 141N, and the northern jet is weaker during the instability as APE is released. Strong anticyclones second season. This is due to the presence of are then present in the northeastern Caribbean in anticyclones in the northeastern Caribbean, which November and December (Fig. 9). MKE increases sustain eastward currents just south of Hispanõla at the end of September as the anticyclonic and Puerto Rico. circulation, partly driven by the freshwater plume ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1465

Fig. 8. Seasonal average of the surface velocity ﬁeld of the MICOM simulation (cm/s). Vectors are plotted every third grid point. Gray areas show the topography mask of the model. (a) January–July and (b) August–December. ARTICLE IN PRESS 1466 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

20 0.5 19 0.45 18 0.4 17 0.35 16 0.3 15 0.25

Latitude 14 0.2 13 0.15 12 0.1 11 0.05

Fig. 10. Time series in J/m2, from the MICOM simulation, of the 20 0.5 spatially and vertically integrated kinetic energy of the mean ﬂow (solid line), of the eddy kinetic energy (dashed-dotted line) 0.45 19 calculated as the departure of the temporal average of the ﬂow, of 18 0.4 available potential energy (dashed line) calculated as the departure from the temporal average of layer thickness and of 17 0.35 the salinity (short dashed line) spatially averaged in the box 16 0.3 61–701W, 13–181N. 15 0.25

Latitude 14 0.2 13 0.15 discretized in coastline and terrain-following curvi- 12 0.1 linear coordinates (s-coordinates model) and has appropriate boundary conditions for an irregular 11 0.05 solid bottom and coastline, free upper surface, and

open-ocean sides away from the coastline. The Longitude boundary conditions include the forcing influences of surface wind stress, heat and water fluxes, coastal Fig. 9. Map of the number of anticyclonic eddies per month in the MICOM simulation. Eddies are counted as they pass each river inflow, bottom drag, open-ocean outgoing grid point. (a) January–July and (b) August–December. wave radiation and nudging towards the specified basin-scale circulation from the climatology. For (Section 3.2), increases the mean speed of the center further information on the capability of the system jet and yields a strong eastward jet just south of the see Shchepetkin and McWilliams (1998, 2003, 2004) islands (Fig. 7). Therefore, the presence of numer- and Marchesiello et al. (2001, 2003). ous anticyclones during August–December is likely The model domain is the northeastern Caribbean to result from the instability fed by the freshwater Sea and the southwestern Tropical Atlantic plume’s APE and EKE resulting from the southern (141–231N, 711–621W), the horizontal resolution is halocline front. 1/181 (6 km), and the model has 25 layers in the In the next subsection we isolate the effect of the vertical. The ocean boundaries were treated as open. salinity plume to understand its role in the circula- Temperature and salinity were relaxed toward their tion in the northeastern Caribbean without the monthly varying climatological values of the Levi- influence of the vorticity shed by the NBC, which is tus climatology (Da Silva et al., 1994). Surface seasonally accounted for in the MICOM simulation fluxes were extracted from the Comprehensive (Garraffo et al., 2003), in agreement with the Ocean Atmosphere Data Set (COADS) climatology. observations (Frantantoni and Richardson, 2006). After a 6-month spinup, the model reached its energetic equilibrium. The model was integrated for 3.2. Regional simulation of the Caribbean current another year, which was used in this analysis. As in the MICOM simulation the ROMS simulation The northern branch (141N) of the Caribbean exhibits the same periods of activities. In particular, Current was studied using the ROMS, which is as the salinity decreases during the summer, the ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1467

Fig. 11. Time-series from the ROMS simulation of the salinity (solid line) and the zonally averaged zonal velocity (m/s) in the surface layer between 621 and 711W from July of the ﬁrst model year to January of the second model year. Dashed contours show the zero velocity. Blue color represents faster westward velocity and orange eastward. The combination of orange in the north and blue in the south (October–December) represents anticyclonic circulation.

Caribbean Current north of 141 is intensified, and temperature of the plume (Ffield, 2005), which the number of anticyclones increases during the fall. decreases its density, must contribute to the We present in Fig. 11 a time series of zonally strengthening of the central jet of the Caribbean averaged flow in the surface layer and of the salinity Current. As seen in the drifter data, the central jet between 621 and 711W. As the salinity decreases, an maximum speed is increased by 10–20 cm/s (Fig. 11) anticyclonic circulation is established along with an during the presence of the freshwater plume. increase of the westward speed of the center jet at The phase diagram of the meridional velocities at 151N (blue) of the Caribbean Current. The simula- 171N is shown in Fig. 13. Between 651 and 681W, tion shows very well the arrival of the freshwater two bands of northward velocities (yellow–red) plume in the northeastern Caribbean in Septem- move westward and surround a band of southward ber–November in agreement with both observations velocities (blue) in July, August, and September. and the MICOM simulation. This corresponds to the arrival of the plume front In order to understand the dynamical effect of the which is an anticylonic anomaly (Fig. 12b), yielding freshwater tongue, the PVA is calculated as the northward velocities on its leading edge. It is difference in PV between the other months (August, followed by a cyclone in September seen at 671W September, October, November, December) and (Fig. 12a). Meanders of positive PVA at the July. The arrival of the pool of negative PVA into southern edge of the negative PVA (Fig. 12b) the basin in August (blue patch, Fig. 12b) is caused suggest that the Caribbean jet at 14–151N becomes by the freshwater plume. The negative PVA then unstable as shown by the MICOM simulation drives the anticyclonic circulation (Fig. 12a). In the energy budget (Fig. 10) in September–October. southern part of the domain, the negative PVA pool Cyclones are then formed on the southern side of generates a PV front where the jet is intensified. the anticyclonic plume and are adevcted north- Finally, as the salinity increases (December) the westward (Fig. 12a, September, October, Novem- PVA is reduced. Hernandez-Guerra and Joyce ber, December). The anticyclonic circulation (2000) found that the center jet at 141N was at the extends farther west to 711W in November southern edge of the freshwater plume and was (Fig. 12a). The interaction of the eastward jet with enhanced geostrophically by the large salinity the shelves of the Greater Antilles generates gradient there. Therefore, from the ROMS simula- cyclones by friction in November and December tion, the low-salinity plume contributes to the (Fig. 12a; D’Asaro, 1988). The cyclones interact intensification of the central jet of the Caribbean with the anticyclonic circulation and break it into Current and generates an anticyclonic circulation smaller anticyclones in November and December, through the negative PVA pool. The warmer between 671 and 711W(Fig. 12a and b, November, ARTICLE IN PRESS 1468 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

Aug Sep

Oct Nov

Dec

Fig. 12a. ROMS simulation snapshots of current (m/s) vectors and salinity (blue indicates freshest water) in m/s. From top to bottom and left to right August–December. The PVA is calculated as the PV difference between July and the other months.

December). Some of the cyclones on the eastern side (Fig. 13). This diagram shows a split in the cyclone of the anticyclonic circulation are advected east- advection at 671W. West of 671W, cyclones are ward, between 621 and 651W during the same period advected westward; east of 671W cyclones are ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1469

Fig. 12b. ROMS simulation snapshots of potential vorticity in m/s. From top to bottom and left to right August–December. The PVA is calculated as the PV difference between July and the other months. ARTICLE IN PRESS 1470 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

the Caribbean current at 141N as seen in the numerical simulations. Indeed, a few months after the peak seasonal rain in the Amazon and Orinoco river basins, a freshwater plume extends northwestward across the eastern Caribbean Sea, starting in July–August and continuing until October– November (Figs. 2 and 4; Muller-Karger et al., 1988; Hernandez-Guerra and Joyce, 2000; Hellwe- ger and Gordon, 2002; Corredor et al, 2004). Two inflows of freshwater enter into the Car- ibbean Sea. The first is south of 121N, where the Orinoco plume, some NBC water, and some NBC ring water enter the Caribbean through Grenada Passage and where the swiftest currents of the Caribbean Current are observed (Fig. 5). The Fig. 13. Phase diagram of meridional velocities at 171N (m/s) second inflow of freshwater is between 141 and from the ROMS simulation. Dashed contours show zero velocity. 181N, partly provided by NBC rings, which stall Yellow–red indicates northward velocities and blue–green south- and decay east of the Lesser Antilles, and partly by ward velocities. the North Equatorial Current, which advects water westward. The main contribution of Amazon water advected eastward. Similar cyclones were observed by rings is during the summer and fall when the by surface drifters to form as Caribbean antic- most of the NBC retroflects into the countercurrent. yclones impinged on the northern Caribbean During August–December, the season of maximum Islands (Richardson, 2005). In summary, Fig. 13 freshwater flux into the Caribbean Sea, the number shows the increase of eddy activity starting at the of drifter anticyclonic looper days per month end of September near 641W and continuing until increases (Fig. 3), which suggests a dynamical link February in the west of the basin (not shown). The with the freshwater input. maximum of the eddy activity, which occurs in To understand this link we used the ECMWF November, follows the arrival of the salinity daily wind-forced high-resolution MICOM simula- minimum in October (Fig. 11). tion, which simulates the triple-jet structure of the Caribbean Current, its variability seen in the drifter 4. Summary and conclusions observations, and the contribution of NBC rings to the freshwater input between 141 and 181N. North The drifter velocities show that the eastern of 131N, a salinity decrease in the MICOM Caribbean Current (61–641W) consists of three jets simulation is associated with an increase of EKE flowing westward at 11.51,141 and 16.81N year and APE, which is later released to EKE (Fig. 10). around. The speeds of the three jets decrease with The center jet of the Caribbean Current intensifies increasing latitude, the southern jet being the as MKE increases and the number of anticyclones fastest. From August to December the same three doubles over the August–December period as jets are observed in the mid-Caribbean (65–701W), compared to January–July. The increase in EKE but during January–July the two southern jets and APE suggests that the buoyant plume is a (11.51,141N) in the east merge to become a single potential vorticity anomaly. The resulting PV jet (12.51N) in the mid-Caribbean. The variability of gradients (horizontal and vertical) could result in speed and structure of the mid-Caribbean Current both barotropic and baroclinic instabilities, which jet is connected to the arrival of the freshwater flux lead to the formation of cyclonic and anticyclonic from the Orinoco and Amazon rivers, which eddies. increases the speed of the mid-Caribbean jet. The It is commonly suggested that instability of the arrival of the freshwater coincides with an increase Caribbean current is driven by the potential of the sea surface temperature (Ffield, 2005). The vorticity fluxes of NBC rings through the island combination of relatively low salinity and warm passages (e.g. Murphy et al., 1999; Carton and temperatures contributes to a sharpening of the Chao, 1999; Johns et al., 2002; Centurioni and density front south of the tongue, which intensifies Niiler, 2003). In order to discriminate the role of the ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1471 freshwater plume from the influence of the vorticity Acknowledgments shed by NBC rings through the Lesser Antilles passages, we studied the circulation in the north- The study was supported by National Science eastern Caribbean influenced only by the freshwater Foundation Grants OCE 03-271808 and OCE using the ROMS model forced by the Levitus 01-36477. The authors thank Chris Wooding for climatology. As the freshwater plume develops a processing the drifter data and creating some of the salinity front at its southern edge, the anticyclonic figures and Dr. M.P. Bacon for careful reading and flow is enhanced and sustained by the negative PVA editing of the manuscript. Terry Joyce discussed of the freshwater plume: the center jet at 141Nis with one of us (PLR) the westward jet associated intensified (Figs. 11,12). It becomes unstable and with the Orinoco River plume, which stimulated generates cyclones on the southern side of the initial interest in this subject. The authors thank the freshwater plume in October–December that are anonymous reviewers for their careful reading and advected northward by the leading edge of the help in improving the manuscript. anticyclonic circulation. The latter could contribute to the northwestward drift of the freshwater plume. The friction of the anticyclonic flow on the References shelves of the Greater Antilles and Virgin Islands generates both westward and eastward propagating Baums, I.B., Paris, C., Che´rubin, L.M., 2006. A dynamical filter cyclones with a separation line near 671W(Fig. 13). in larval dispersal in a reef building coral. Limnology and Oceanography 51 (5), 1969–1981. The cyclones interact with the anticyclonic flow Bleck, R., Chassignet, E.P., 1994. Simulating the oceanic causing it to break into smaller anticyclones. circulation with isopycnic coordinates models. In: Majumdar, Eddy activity is significantly increased after the S.K., et al. (Eds.), The Oceans: Physiochemical Dynamics and arrival of the freshwater plume. These results Resources. Pennsylvania Academy of Science, Harrisburg, suggest that the freshwater plume in the eastern pp. 17–39. Bleck, R., Rooth, C., Hu, D., Smith, L.T., 1992. Salinity-driven Caribbean, independent of the vorticity of NBC transients in a wind-and thermohaline-forced isopycnic rings, is a main source of variability of the coordinate model of the North Atlantic. Journal of Physical Caribbean Current, in addition to the wind and Oceanography 22, 1486–1505. thermohaline-driven flows. Candela, J., Beardsley, R.C., Limeburner, R., 1992. Separation of As noted around Barbados Island by Cowen et al. tidal and subtidal currents in ship mounted acoustic Doppler current profiler (ADCP) observations. Journal of Geophysical (2003), the freshwater flux from NBC rings exerts a Research 97 (C1), 769–788. strong forcing on the reef ecosystem. In particular, Carton, J.A., Chao, Y., 1999. Caribbean Sea eddies inferred from Cowen et al. (2003) observed a strong biological TOPEX/POSEIDON altimetry and 1/61 Atlantic Ocean response of the system as measured by recruitment model simulation. Journal of Geophysical Research 104 of coral reef fishes. During the presence of rings, (C4), 7743–7752. Centurioni, L.R., Niiler, P.P., 2003. On the surface currents of some events rapidly advected larval fish away from the Caribbean Sea. Geophysical Research Letters 30 (6), 1279. the island, resulting in a failure of larval settlement. Chassignet, E.P., Garraffoo, Z.D., 2001. Viscosity parameteriza- Under other conditions larval retention was en- tion and the Gulf Stream separation. From Stirring to Mixing hanced and was followed by a settlement pulse in a Stratified Ocean: Proceedings of the ‘Aha Huliko’a (see also Baums et al., 2006). Moreover, changes in Hawaiin Winter Workshop, Honolulu, HI University of Hawaiian at Manoa, pp. 37–41. the vertical distribution of fish larvae were observed Corredor, J.E., Morell, J.M., Lopez, J.M., Capella, J.E., simultaneous with the intrusion of freshwater. Armstrong, R.A., 2004. Cyclonic eddy entrains Orinoco Larval fish encountering fresh NBC ring water grew River plume in eastern Caribbean. Eos, Transactions, more slowly and had longer larval periods, both American Geophysical Union 85(20), 197, 201–202. potentially reducing survival and ultimately, recruit- Cowen, R.K., Castro, L.R., 1994. Relation of Coral reef fish larval distributions to island scale circulation around Barba- ment success. Therefore, results by Cowen et al. dos, West Indies. Bulletin of Marine Science 54, 228–244. (2003) demonstrate that the freshwater plume Cowen, R.K., Sponaugle, S., Paris, C.B., Fortuna, J.L., Lwiza, interferes with the island-scale flow dynamics and K.M.M., Dorsey, S., 2003. Impact of North Brazil Current injects considerable variability in the local reef rings on local circulation and coral reef fish recruitment to ecosystems. This study contributes to a better Barbados, West Indies. In: Gon˜i, G.J. (Ed.), Interhemispheric Water Exchange. Elsevier Oceanographic Series, pp. 443–462, understanding of the seasonal variability of the Chapter 17. Caribbean Current and its potential influence in the D’Asaro, E.A., 1988. Generation of submesoscale vortices: a new Caribbean Sea coastal ecosystems. mechanism. Journal of Geophysical Research 93, 6685–6693. ARTICLE IN PRESS 1472 L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473

Da Silva, A.M., C Young, C., Levitus, S., 1994. Atlas of Surface space and S-PALACE floats. Deep-Sea Research II 51, Marine Data 1994, vol. 1, Algorithms and Procedures. 1151–1171. NOAA Atlas NESDIS 6, US Department of Commerce, Johns, W.E., Lee, T.N., Schott, F.A., Zantopp, R.J., Evans, NOAA, NESDIS, USA, 74pp. R.H., 1990. The North Brazil Current retroflection: seasonal Dessier, A., Donguy, J.R., 1994. The sea surface salinity in the structure and eddy variability. Journal of Geophysical Tropical Atlantic between 101S and 301N—seasonal and Research 95 (C12), 22103–22120. interannual variations (1977–1989). Deep-Sea Research I 41, Johns, W.E., Lee, T.N., Beardsley, R., Candela, J., Castro, B., 81–100. 1998. Annual cycle and variability of the North Brazil Ffield, A., 2005. North Brazil current rings viewed by TRMM Current. Journal of Physical Oceanography 28 (1), 103–128. Microwave Imager SST and the influence of the Amazon Johns, E., Wilson, W.D., Molinari, R., 1999. Direct observations Plume. Deep-Sea Research I 52, 137–160. of velocity and transport in the passages between the Intra- Fratantoni, D.M., 2001. North Atlantic surface circulation Americas Sea and the Atlantic Ocean, 1984–1996. Journal of during the 1990s observed with satellite-tracked drifters. Geophysical Research 104 (C11), 25,805–25,820. Journal of Geophysical Research 106 (C10), 22067–22093. Johns, W.E., Townsend, T.L., Fratantoni, D.M., Wilson, W.D., Frantantoni, D.M., Richardson, P.L., 2006. The evolution and 2002. On the Atlantic inflow into the Caribbean Sea. Deep- demise of North Brazil Current rings. Journal of Physical Sea Research I 49, 211–243. Oceanography 36, 1241–1264. Johns, W.E., Zantopp, R.J., Gon˜i, G.J., 2003. Cross-gyre Fratantoni, D.M., Glickson, D.A., 2002. North Brazil Current transport by North Brazil Current rings. In: Gon˜i, G.J., ring generation and evolution observed with Sea WiFS. Malanotte-Rizzoli, P. (Eds.), Interhemispheric Water Ex- Journal of Physical Oceanography 32, 1058–1074. change in the Atlantic Ocean. Elsevier Oceanographic Series Fratantoni, D.M., Johns, W.E., L Townsend, T., 1995. Rings of 68, Elsevier, Amsterdam, pp. 411–441. the North Brazil Current: their structure and behavior Kelly, P.S., Lwiza, K.M.M., Cowen, R.K., 2000. Low-salinity inferred from observations and a numerical simulation. pools at Barbados, West Indies: their origin, frequency, and Journal of Geophysical Research 100 (C6), 10633–10654. variability. Journal of Geophysical Research 105 (C8), Garraffo, Z.D., Mariano, A.J., Griffa, A., Veneziani, C., 19,699–19,708. Chassignet, E.P., 2001. Lagrangian data in a high-resolution Levitus, S., 1982. Climatological Atlas of the World Ocean, numerical simulation of the North Atlantic I, Comparison NOAA/ERL GFDL Professional Paper 13, Princeton, NJ, with in situ drifter data. Journal of Marine Systems 29, 173pp (NTIS PB83-184093). 157–176. Marchesiello, P., McWilliams, J.C., Shchepetkin, A., 2001. Open Garraffo, Z.D., Johns, W.E., Chassignet, E.P., Gon˜i, G.J., 2003. boundary conditions for long-term integration of regional North Brazil Current rings and transport of southern waters oceanic models. Ocean Modelling 3, 1–20. in a high resolution numerical simulation of the North Marchesiello, P., McWilliams, J.C., Shchepetkin, A., 2003. Atlantic. In: Gon˜i, G.J., Malanotte-Rizzoli, P. (Eds.), Equilibrium structure and dynamics of the California Current Interhemispheric Water Exchange in the Atlantic Ocean. System. Journal of Physical Oceanography 33, 753–783. Elsevier Oceanographic Series 68, Elsevier, Amsterdam, Mayer, D.A., Weisberg, R.H., 1993. A description of COADS pp. 375–409. surface meteorological fields and the implied Sverdrup Garzoli, S.L., Ffield, A., Yao, Q., 2003. North Brazil Current transports for the Altantic Ocean from 301Sto601N. Journal rings and the variability in the latitude of retroflection. In: of Physical Oceanography 23, 2201–2221. Gon˜i, G.J., Malanotte-Rizzoli, P. (Eds.), Interhemispheric Molinari, R.L., 2004. Annual and decadal variability in the Water Exchange in the Atlantic Ocean. Elsevier Oceano- western subtropical North Atlantic: signal characteristics and graphic Series 68, Elsevier, Amsterdam, pp. 357–375. sampling methodologies. Progress in Oceanography 62 (1), Gon˜i, G.J., Johns, W.E., 2001. A census of North Brazil Current 33–66. rings observed from TOPEX/POSEIDON altimetry: Muller-Karger, F.E., McClain, C.R., Richardson, P.L., 1988. 1992–1998. Geophysical Research Letters 28 (1), 1–4. The dispersal of the Amazon’s water. Nature 333, 56–59. Gon˜i, G.J., Johns, W.E., 2003. Synoptic study of warm rings in Muller-Karger, F.E., R McClain, C., Fisher, T.R., Esaias, W.E., the North Brazil Current retroflection region using satellite Varela, R., 1989. Pigment distribution in the Caribbean Sea: altimetry. In: Gon˜i, G.J., Malanotte-Rizzoli, P. (Eds.), Observations from space. Progress in Oceanography 23, Interhemispheric Water Exchange in the Atlantic Ocean. 23–64. Elsevier Oceanographic Series 68, Elsevier, Amsterdam, Murphy, S.J., Hurlburt, H.E., O’Brien, J.J., 1999. The con- pp. 335–356. nectivity of eddy variability in the Caribbean Sea, the Gulf of Hamilton, P., Larsen, J.C., Leaman, K.D., et al., 2005. Mexico, and the Atlantic Ocean. Journal of Geophysical Transports through the Straits of Florida. Journal of Physical Research 98 (C5), 8389–8394. Oceanography 35 (3), 308–322. Niiler, P.P., Sybrandy, A.S., Bi, K., Poulain, P.M., Bitterman, Hellweger, F.L., Gordon, A.L., 2002. Tracing Amazon River D., 1995. Measurements of the water-following characteristics water into the Caribbean Sea. Journal of Marine Research 60 of Tristar and Holey-sock drifters. Deep-Sea Research 42, (4), 537–549. 1951–1964. Hernandez-Guerra, A., Joyce, T.M., 2000. Water masses and Perry, G.D., Duffy, P.B., Miller, N.L., 1996. An extended data circulation in the surface layers of the Caribbean at 661W. set of river discharges for validation of general circulation Geophysical Research Letters 27 (21), 3497–3500. models. Journal of Geophysical Research 101, 21339–21349. Hu, C., Montgomery, E.T., Schmitt, R.W., Muller-Karger, F.E., Richardson, P.L., 2005. Caribbean Current and eddies as 2004. The dispersal of the Amazon and Orinoco River water observed by surface drifters. Deep-Sea Research II 52, in the tropical Atlantic and Caribbean Sea: observation from 429–463. ARTICLE IN PRESS L.M. Che´rubin, P.L. Richardson / Deep-Sea Research I 54 (2007) 1451–1473 1473

Richardson, W.S., Schmitz Jr., W.J., Niiler, P.P., 1969. The Shchepetkin, A.F., McWilliams, J.C., 1998. Quasi-monotone velocity structure of the Florida Current from the Straits of advection schemes based on explicit locally adaptive dissipa- Florida to Cape Fear. Deep-Sea Research 16 (Suppl), tion. Monthly Weather Review 126, 1541–1580. 225–234. Shchepetkin, A.F., McWilliams, J.C., 2003. A method for Romanou, A., Chassignet, E.P., Sturges, W., 2004. The Gulf of computing horizontal pressure-gradient force in an ocean Mexico circulation within a high-resolution numerical simu- model with a non-aligned vertical coordinate. Journal of lation of the North Atlantic Ocean. Journal of Geophysical Geophysical Research 108 (C3), 3090. Research 109. Shchepetkin, A., McWilliams, J.C., 2005. The regional oceanic Schmitz, W.J., McCartney, M.S., 1993. On the North Atlantic modeling system: a split-explicit, free-surface, topography- circulation. Reviews of Geophysics 31 (1), 29–49. following-coordinate ocean model. Ocean Modelling 9, Schmitz, W.J., Richardson, P.L., 1991. On the sources of the 347–404. Florida Current. Deep-Sea Research 38 (Suppl. 1), 379–409. Simmons, H.L., Nof, D., 2002. The squeezing of Eddies through Schott, F.A., Lee, T.N., Zantopp, R., 1988. Variability of gaps. Journal of Physical Oceanography 32 (1), 314–335. structure and transport of the Florida Current in the period Sybrandy, A.L., Niiler, P.P., 1991. WOCE/TOGA Lagrangian range of days to seasonal. Journal of Physical Oceanography drifter construction manual. SIO ref. 91/6, WOCE Rep. 63. 18, 1209–1230. Scripps Institution of Oceanography, La Jolla, CA, 58pp.